Author: Kim Roper Jensen
Date: 13:58:15 07/22/04
Go up one level in this thread
On July 22, 2004 at 16:41:08, Steven Edwards wrote: Hi Steven Im really looking forward to see your creation play, please say when you will play the first games on fics or icc. And a little question, how does Symbolic implement evaluation ? With regards Kim >Symbolic: Status report 2004.07.22 > >Work continues on the CTXboard class, and now most of the xboard protocol >version 2 commands are fully handled. Everything needed for automated matches >is now operating; however, there are still several commands used for interactive >play that are not yet hooked up to Symbolic's toolkit. Some of these, like >"remove" and "undo" won't be too hard to implement. Others, like "sd" and >"edit" aren't really worth the effort and I'll probably leave them out of the >CTXboard implementation. > >The toolkit allows for a time control with three segments: each segment >specifies a move count, a time allocation, and an increment. One, two, or three >segments may be present in a time control. If a sudden death (move count equals >zero) segment is present, there can be only one and it must be the last segment. > If the last segment isn't a sudden death segment, then it repeats in play until >the game ends. I designed the CTTimeControl class and the CTTCSegment class to >support just about every tournament I've seen advertised; none that I recall had >more than three time control segments. > >However, there is still some work to be done in the toolkit for actual >pre-search time allocation. At the moment it only understands about sudden >death single segment time controls, and it allocates nominal search time by >dividing the remaining time by thirty. Simple, but it seems to work okay. Now >that automated matches are supported I can better test non sudden death >controls, so that topic is high on the priority list. Alas, Crafty only >supports a two segment time control and xboard can handle only a single segment >time control, so testing of multiple segment controls will have to wait. > >I've been looking at the zippy portion of xboard for enabling ICS play. My >working assumption is that xboard can handle the details as long as I pass it >the correct command line options and a working ICS login script file. Perhaps I >will get around to testing this next month. > >To speed up the automated matches, I've added the CTResigner class and hooked it >up to CTXboard so that the engine can give up the struggle in a gentlemanly way >when things get too rough. The default value of "too rough" in CTResigner will >trigger engine resignation when faced with a negative score of at least a rook >for five consecutive searches. > >While there hasn't been much in the way of Lisp work in the past couple of >months, nearly all of the work on toolkit issues indirectly supports or extends >Symbolic's Lisp program source and so it will all pay off (I hope) when the real >Symbolic starts playing full games intelligently.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.